What happened
Volantis closed an $88 million Series A round on Oct. 1, 2026, bringing total disclosed funding to $97 million. The round was co‑led by Lachy Groom and Abstract Ventures and included investors such as John Doerr, VXI Capital, Triatomic and Susa Ventures.
According to a PR Newswire release and the company’s blog, Volantis raised $88 million in a Series A round announced on Oct. 1, 2026. The round was co‑led by Lachy Groom and Abstract Ventures, with participation from John Doerr, VXI Capital, Triatomic, Susa Ventures and several angels. The funding brings the startup’s total disclosed capital to $97 million since its 2022 founding by CEO Tapa Ghosh and CTO Roy Meade.
Volantis describes its technology as a VCSEL‑based photonic interconnect that can link AI compute dies to memory pools over distances of 200 mm or more—roughly 100 × the reach of traditional copper traces. The company claims this could allow aggregation of up to 100 × more memory chiplets into a single high‑bandwidth pool, though no independent performance data have been released.
The announcement also referenced long‑term goals such as surrounding a GPU with about 220 memory chips, supporting models beyond 20 trillion parameters, and delivering roughly 10,000 tokens per second per user. These figures are reported targets, not verified specifications.
Source details: shattered.io ↗
Why it matters
The raise underscores growing investor confidence that memory bandwidth, not raw compute, will be the next bottleneck for large‑scale AI models. By promising optical links that can reach 100 × farther than copper, Volantis hopes to enable far larger memory pools per chip, potentially easing the chronic HBM shortage and lowering the cost of high‑performance hardware.
Memory bandwidth has become the primary constraint for on trillion‑parameter models. Current solutions rely on high‑bandwidth memory (HBM) stacked directly on the GPU die, but HBM supply is tight and its electrical reach is limited by copper trace physics. Volantis’s optical approach could sidestep the “shoreline” limit, potentially unlocking larger memory capacities without redesigning the entire GPU package.
If the technology matures, it would give cloud providers a new lever to negotiate better pricing and performance from GPU vendors, especially as HBM shortages push up hardware costs. A successful photonic link could also influence the roadmaps of Nvidia and AMD, which have heavily invested in HBM‑centric designs.
The investor roster—including prominent AI‑focused angels and venture partners—signals that capital markets view the memory‑bandwidth challenge as a strategic inflection point for the AI hardware ecosystem.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Which component of an AI application is the machine-learning model itself?
What to watch next
Key indicators will be Volantis’s prototype milestones, any third‑party data on bandwidth or energy per bit, and whether cloud providers or GPU vendors commit to pilot deployments before the company’s 2027 target for its first engine.
Prototype demonstrations that include measured bandwidth, latency and energy per bit, ideally validated by an independent lab.
Announcements of pilot programs or design‑win agreements with major cloud providers (e.g., AWS, Azure, Google Cloud) or GPU manufacturers.
Updates on the company’s timeline for a 2027 engine release, and any revisions to the claimed memory‑chip aggregation numbers.
Competitive activity from other photonic interconnect firms such as Ayar Labs, Lightmatter, and emerging silicon‑photonic startups, which could affect Volantis’s market positioning.